Rare Variant Analysis of Human and Rodent Obesity Genes in Individuals with Severe Childhood Obesity

Abstract

Obesity is a genetically heterogeneous disorder. Using targeted and whole-exome sequencing, we studied 32 human and 87 rodent obesity genes in 2,548 severely obese children and 1,117 controls. We identified 52 variants contributing to obesity in 2% of cases including multiple novel variants in GNAS, which were sometimes found with accelerated growth rather than short stature as described previously. Nominally significant associations were found for rare functional variants in BBS1, BBS9, GNAS, MKKS, CLOCK and ANGPTL6. The p.S284X variant in ANGPTL6 drives the association signal (rs201622589, MAF~0.1%, odds ratio = 10.13, p-value = 0.042) and results in complete loss of secretion in cells. Further analysis including additional case-control studies and population controls (N = 260,642) did not support association of this variant with obesity (odds ratio = 2.34, p-value = 2.59 × 10^-3), highlighting the challenges of testing rare variant associations and the need for very large sample sizes. Further validation in cohorts with severe obesity and engineering the variants in model organisms will be needed to explore whether human variants in ANGPTL6 and other genes that lead to obesity when deleted in mice, do contribute to obesity. Such studies may yield druggable targets for weight loss therapies.

Figure 1

Identification of clinically-associated variants in known human obesity genes. Schematic outlining the analysis strategy.

Figure 2

Pedigrees of patients with clinically associated variants contributing to obesity Co-segregation of variants is shown where male (squares) and female (circles) family members consented to genotyping. Heterozygous (filled) and wild-type (empty) mutation carriers are indicated; in some cases, individuals were not available for genotyping (grey). Body mass index (BMI) (>27 kg/m² = overweight;>30 kg/m² = obesity) for adults and BMI standard deviation scores (BMI sds) for children are shown where data was available. (A) Obesity alone; (B) Obesity and Delay. MC4R mutations were excluded by prior Sanger sequencing of all individuals undergoing whole exome sequencing and the majority (1744/1811) of individuals undergoing targeted resequencing. The finding of four MC4R mutations in the remaining 82 individuals is in keeping with the prevalence of heterozygous MC4R mutations in this cohort as reported previously.

Figure 3

Structural model of variants identified in GNAS. (A) Structure of the active state ternary protein complex of G-protein coupled receptor (GPCR) beta 2-adrenergic receptor (ADRB2), and the nucleotide-free Gs heterotrimer, composed of Gs-α(GNAS), Gs-β, and Gs-γ subunits (based on Rasmussen et al., pdb file: 3sn6). The two major domains of GNAS are noted, α-helical domain and Ras-like GTPase domain. The principal interactions between GNAS and GPCRs involve the amino- and carboxy-terminal α-helices. The Ras-like GTPase domain contains most of the catalytic residues necessary for GTP hydrolysis, as well as the Gβγ and effector binding regions (switch regions I, II and III), which change confirmation upon binding to GTP or GDP). (B) Structure of the complex of Gs-alpha with the catalytic domains of mammalian adenylyl cyclase (based on Tesmer et al., pdb file: 1azs). Variant residues (purple), and the different components of the complex are highlighted.

Figure 4

Functional characterisation of ANGPTL6 variants location and functional effect (A) Human ANGPTL6 protein and location of the genetic variants. Schematic of the human ANGPTL6 protein with the three recognisable domains: signal peptide (SP) in green, coiled-coil domain (CCD) in purple, and fibrinogen-like domain (FLD) in yellow. As all genetic variants identified are located predominantly in the FLD, the partial tertiary structure of the C-terminal protein portion (amino acids 135–470), based on previously solved fibrinogen structure (1lwu) is presented. The single variant seen in controls only is shown in blue, variants identified in cases are in purple, the S284X variant is starred. (B) Functional characterisation of ANGPTL6 variants. Cells were transiently transfected with constructs encoding wild-type (WT) or mutant ANGPTL6; levels of protein secretion into the media were measured by ELISA; means+/− standard deviation (SD) for experiments performed in triplicate are shown (results were confirmed by Western blotting; data not shown). Statistical significance was measured using unpaired T-test with Welch’s correction using the GraphPad Prism software. *p < 0.05; ***p < 0.001.

Figure 5

Tiered analysis of obesity candidate genes sets. ORs (bars) and 95% (dotted vertical lines) confidence intervals for each of the six gene tiers and across all 119 genes combined each for eight different filtering scenarios by MAF (Rare or Novel) and functional prediction (synonymous - Synon, functional - Func, functional with at least 1 deleterious consequence prediction – Func, 1 + del, and functional with all deleterious consequence predictions - Func, all del).